Tools I developed to make my preperation easy

dapu

@gpsias You seem to be from IT background. I am planning to go with EE or Philosophy..not decided yet ..Which optional you going for....are you too first timer?

I'm a civil Engineer. This is just out of sheer passion.
Will be releasing a video and a blog in a week explaining what these tools can do.

@gpsias :O how you know that? I tried but that project is not related to this app. I didn't found corpus for Hindi and acoustic models were pretty lame too. I have stopped working on that as I was not getting time for exams. And it was actually not an app, but a program intended for IoT device.

I just tried english. Not sure about hindi. Any speech to text to gain decent accuracy should be linked with internet. Else it will be constrained and it needs to be trained. but Nuance dragon is good except that It cant recognise new words like from our Ancient history and culture .

That is the main reason for developing this tool. It can recognize even such words if they are well known. Now need to give it any context or reference.

That is great. looking for update in future .

For English also, there are variety of accents in Indian English as compared to US English. The acoustic model of US-english if used here in India would lead to less accuracy, infact horrible. Naunce one is extremely costly, not sure if one can afford it even. The google's Speech API is free only for limited time and will cost 6 cents per query I guess afterwards. Best is to train own CMU Sphinx model and launch. I was hoping that you used the same, would surely love to see how well is recognition in your tool. I would love to learn more on this.

renaissance

Itna kyun thak riya hai

gpsias

@gpsias You seem to be from IT background. I am planning to go with EE or Philosophy..not decided yet ..Which optional you going for....are you too first timer?

I'm a civil Engineer. This is just out of sheer passion.
Will be releasing a video and a blog in a week explaining what these tools can do.

@gpsias :O how you know that? I tried but that project is not related to this app. I didn't found corpus for Hindi and acoustic models were pretty lame too. I have stopped working on that as I was not getting time for exams. And it was actually not an app, but a program intended for IoT device.

I just tried english. Not sure about hindi. Any speech to text to gain decent accuracy should be linked with internet. Else it will be constrained and it needs to be trained. but Nuance dragon is good except that It cant recognise new words like from our Ancient history and culture .

That is the main reason for developing this tool. It can recognize even such words if they are well known. Now need to give it any context or reference.

That is great. looking for update in future .

For English also, there are variety of accents in Indian English as compared to US English. The acoustic model of US-english if used here in India would lead to less accuracy, infact horrible. Naunce one is extremely costly, not sure if one can afford it even. The google's Speech API is free only for limited time and will cost 6 cents per query I guess afterwards. Best is to train own CMU Sphinx model and launch. I was hoping that you used the same, would surely love to see how well is recognition in your tool. I would love to learn more on this.

This is working with indian english and we have to pass that parameter before initializing.
And google speech API which I have used, is leveraging the browser's capability to Speech to text and it is most unlikely to become paid in near future. But until then, yeah it is working well with our desi english. No need of hifi US english.

And CMU Sphinx needs to be trained with lot of data before putting it to use and it again has the same old problem of not being able to detect new words. This more or less is like a open source alternative of nuance engine, isn't it ? . . . Correct me if I'm wrong .

dapu

@gpsias You seem to be from IT background. I am planning to go with EE or Philosophy..not decided yet ..Which optional you going for....are you too first timer?

I'm a civil Engineer. This is just out of sheer passion.
Will be releasing a video and a blog in a week explaining what these tools can do.

@gpsias :O how you know that? I tried but that project is not related to this app. I didn't found corpus for Hindi and acoustic models were pretty lame too. I have stopped working on that as I was not getting time for exams. And it was actually not an app, but a program intended for IoT device.

I just tried english. Not sure about hindi. Any speech to text to gain decent accuracy should be linked with internet. Else it will be constrained and it needs to be trained. but Nuance dragon is good except that It cant recognise new words like from our Ancient history and culture .

That is the main reason for developing this tool. It can recognize even such words if they are well known. Now need to give it any context or reference.

That is great. looking for update in future .

For English also, there are variety of accents in Indian English as compared to US English. The acoustic model of US-english if used here in India would lead to less accuracy, infact horrible. Naunce one is extremely costly, not sure if one can afford it even. The google's Speech API is free only for limited time and will cost 6 cents per query I guess afterwards. Best is to train own CMU Sphinx model and launch. I was hoping that you used the same, would surely love to see how well is recognition in your tool. I would love to learn more on this.

This is working with indian english and we have to pass that parameter before initializing.
And google speech API which I have used, is leveraging the browser's capability to Speech to text and it is most unlikely to become paid in near future. But until then, yeah it is working well with our desi english. No need of hifi US english.

And CMU Sphinx needs to be trained with lot of data before putting it to use and it again has the same old problem of not being able to detect new words. This more or less is like a open source alternative of nuance engine, isn't it ? . . . Correct me if I'm wrong .

It is actually paid only if you make your app commercial. For personal use they won't even bother you ...This is good way to do, since this is tool for helping all of us. And yea CMU Sphinx is more or less the same like Naunce, but still Naunce is way better than Sphinx toolkit as latter one is meant for research purposes. Whereas the Naunce is reputed and professional TTS & ASR provider.

Google ASR choice is good since everyone today preparing for UPSC will surely have internet. You app must be very fast and lightweight...Please do share asap, we all would love to have it.

gt2015

@gpsias bro please provide the link for the software or tell us how to get/use this software.

messimik

I am new to Speech-to-Text
Can you please guide me what should I do (as in which app should I download) to make notes?

gt2015

I am new to Speech-to-Text
Can you please guide me what should I do (as in which app should I download) to make notes?

If you have any microphone/ headphone with mic connect it to the pc then go to Google and type " Google speech API".

dapu

.

messimik

I am new to Speech-to-Text
Can you please guide me what should I do (as in which app should I download) to make notes?

If you have any microphone/ headphone with mic connect it to the pc then go to Google and type " Google speech API".

Thanks for helping

But 'Speech API - Speech Recognition | Google Cloud Platform' requires Credit Card information which I don't feel safe to provide.


I tried using 'Web Speech API Demonstration'.
It works well.

But it does not detect many words.

gt2015

I am new to Speech-to-Text
Can you please guide me what should I do (as in which app should I download) to make notes?

If you have any microphone/ headphone with mic connect it to the pc then go to Google and type " Google speech API".

Thanks for helping

But 'Speech API - Speech Recognition | Google Cloud Platform' requires Credit Card information which I don't feel safe to provide.


I tried using 'Web Speech API Demonstration'.
It works well.

But it does not detect many words.

Did you selected English language and country India ? If not then after selecting English language select country as India then it will be able to detect almost every word (except some ancient history , culture related words).